Fab engineers save Intel's butt

Although you have not replied to any of my e-mails recently I offer you these one last set of observations. [Dear Chipster, I have replied. There must be something awry, Ed.]

What you are seeing from Intel right now is people clearing off their desks on the way to their next assignments. I mentioned in January that the target specs on the next generation (Tejas) would be 1000 MHz bus (or 1066 depending on divisor) with 2 MB of cache. Since the departing designers do not need to keep these things in their bag of tricks and as they are relatively simple to implement they went ahead and added these already developed features to the production database for the Prescott, leaving the engineers at the fab then capable of generating masks which incorporate these features. The marketing department, who have been taking quite a bit of abuse about the lackluster performance of the Prescott, have immediately jumped on the availability (remember these fab engineers are the ones who saved Marketing's behinds with the extreme edition and L3 cached Xeons and must have went running to marketing as soon as they saw what they had).

Ditto with Potomac, Potomac must already have been taped out which means that the primary design team could then have been released since there will be no follow on work to this processor (tape out should have commenced three months after Nocona cleared production validation which means it should have been completed by now). Since there is only a single design revision look on this to be a very simple implementation of MP (as was Gallatin), 2 MB or 4 MB of L3 cache, 1 MB of L2 cache, 533 MHz bus (going to 800 MHz in an extreme edition), barely working 64 bit technology. Since it costs next to nothing to carry the DP and MP circuitry in the desktop chip look for Intel to simplify the Prescott/Nocona/Potomac lines to two chip variants, with and without L3 cache and then use bond out options to select what it can do ala the Opteron. This then leaves one caretaker design group living in the desktop division to service the full line (one change changes both databases).

More interesting is Bill Gates comments that he expects all servers and most workstations to be 64 bit technology in the near future. If anyone should know it is him.

Chipster

Reviews

Regarding benchmarks, benchmarketing and methodologies, one of the notable blindspots is AV software. AFAIK, a typical Windows PC will have AV software. It's almost compulsory for home/office usage.

So far I haven't seen any benchmarks indicating which CPU architecture is better for AV software alone or popular benchmarks with AV software active. There's a lack of info on whether HT should be ON/OFF or if McAfee is better with P4, Sophos is better with Athlons etc.

I dunno about the reviewer PCs, but in my experience, AV software is endemic on desktops whether or not it is kept up to date, it is still exacting a performance tax on the PC. Nowadays a significant number even have firewall software on.

Due to these omissions, so far the popular windows desktop benchmarks I've seen aren't close to "real world".

Lincoln Yeoah

Apple and Trojan Horses

Nick,

I'm going to assume that your "trojan horse" article is just trying to inflame people. Is it really possible you don't have any idea what you are talking about?

"Trojan horse has penetrated the Mac OSX" Um, no. A Trojan horse doesn't "penetrate" anything - it's invited in. Read up on your Greek epics.

The point is, users are allowed to delete their own files. This idiot downloaded a program he couldn't trust and ran it. ANY program can written for ANY operating system can delete files.

If you run an untrustworthy program, it can do anything you could do manually.

Ah, hell, just read what I wrote at http://www.danshockley.com/weblog.php. Maybe you'll learn something, maybe not. Maybe you don't care - I've read a few articles at inquirer.net before.

Cheers
Daniel A. Shockley

German Hackers

Hi,

"One leading German newspaper, Die Welt said in a page one editorial on Monday there was a strange sense of national pride that a German student had outwitted the world's best computer experts. [Are you sure about this? Ed.]"

They did indeed mention national pride, if not in so many words. See http://www.welt.de/data/2004/05/10/276105.html?search=sasser&searchHILI=1

The relevant part of the article, apart from what has already been quoted, seems to be

"Dass ausgerechnet deutsche Schüler (Pisa!) in der Lage sind, weltweit doch noch für Aufsehen zu sorgen und gegen Microsoft anzutreten: irgendwie faszinierend. Gerüchte kursieren, dass die Geschichte des Niedersachsen längst verkauft ist. Vielleicht an Hamburger Magazine oder an Hollywood? Auch die Computerindustrie würde sich bald mit Jobangeboten melden. Vielleicht ein Trost für die Eltern, die Samstagmorgen den Satz hören mussten: "Ihr Sohn hat die Welt lahm gelegt."

I'm too lazy to translate that verbatim, alas. In a nutshell, it's "fascinating" that German students, who got a bashing regarding their bad education in the Pisa study, nonetheless manage to create worldwide furor and "compete against Microsoft". There are, the paper further reports, rumours that the story of the student has been sold, maybe to Hollywood, maybe to a Hamburg magazine. The IT industry is rumoured to make job offers soon, too.

There, I did translate most of that after all. I find the language unfortunate. "Compete against Microsoft"? It's not a sports event, for heaven's sake. Worms aren't cool, nor are their authors "1337". They're just sad. But hey, maybe you need to be out of diapers to get that.

Thorsten

Hardware Sites and Methodology

Lacking a measure of uncertainty, normally reported as "standard deviation", the numbers published on "hardware review sites" have no credibility from a statistical point of view.

We can only guess if biases or pressures resulted in the numbers published being best-case, worst-case or convenient-case numbers.

When a number like 2200.7 is reported alone, the reader stands no chance to know if this should be interpreted as "never below 2200" or "probably above 1000". If it was reported with a standard deviation of 23 then we do know.

When comparing two products, the easy to use "Students t test" determines if there actually is a credible difference between the two set of tests.

Statistics at this level is not hard. Try the excellent book "Cartoon guide to statistics" by Larry Gonick and Wollcott Smith.

Poul-Henning

AMD still only second best

Mike,

I can't help but to think with all the advantages AMD currently has with Intel in panic mode trying to get a competing roadmap together while AMD keeps churning out better and better technology; that with all of this AMD will still only be second best.

Let me start this out with (in case you post) that I'm an AMD fan boy from the K6 days and I've never purchased an Intel kit in my life. Furthermore I love and admire AMD's engineers and take my hat off to their efforts. Now that's out of the way, let's talk about management and marketing.

Every now and again AMD puts out a really impressive technology, and as the years go on, those occurrences produce processors that are more impressive then the last. AMD concentrates on making their chips smarter, not just faster and from an engineering prospective, I've always thought this to be more noteworthy.

But once the engineers bring forth this idea, management and marketing seems to put a pillow over it trying to smother its life. Everyone already knows the history so I'll try and stick to my latest gripe - AMD64.

We all know AMD can shoot itself in the foot at its every whim and I'm sure it will do so again with the AMD64 technology. We have the Opteron, and Athlon 64 being some of the best silicon produced by any fab. We also have Intel having troubles matching the demand for AMD64 and are trying to regroup their roadmaps to position iAMD64 into the mainstream. So one would think AMD would take full advantage of this and speed past Intel. As history would have it, AMD will pull back on the throttle and just stay competitive thinking why show all our cards when Intel can't keep up anyway. Well with less than 20% of market share I would do everything I could to grab up every last percentage which would in turn allow for more money in the R&D department or other areas of need.

I would push the clock up and tweak every last IPC out of the core. I would put every effort to getting cooler mobile/server chips out of the fab that I could pushing AMD's lead far above Intel. This will also put pressure on those engineers to get dual core out the door. And don't feel bad for them as they very much like the pressure being it helps them think better.

I would get the dual core CPUs out the door on the 90 nano as quick as possible and have those dual cores also implementing Twin Threading Plus which would have multithreading on both cores but allow the core to turn Twin Threading on/off depending on the code being processed to maximize efficiency. This gives you either both cores non Twined, or one core Twined and the other not, or both cores Twined. Of course the marketing guys could have a field day with ‘The Twins' I'm sure ?

I would not stop there, but also put some cash into Linux and having a small team get the user friendly side of Linux into existence. Hire some authors of children's books to rewrite the help files, menus, install guild, and any other language that is 100% geeked so that everyday people can use it. Remember, “An Apple a day keeps the Help Desk away”. Also make sure they have whatever they need to produce code that is very friendly to the current and upcoming AMD64 architectures. This will put needed pressure on Microsoft to get 64bit code or extensions into Windows.

But we all know I'm dreaming and AMD will not throw any punches while Intel is against the ropes and will continue to waste their engineer's talents. The idea of reserve your strength for a future fight has not worked and AMD needs to strike lethal blows when it can. They have the opportunity to really make a dent in the server space and even mobile if they get some cool chips out the door. With Dell having problems in Europe I would find an OEM willing to take the lead and feed them everything they need to make a mark over there. I would have the marketing guys change strategies and make them accountable to intelligent thinking like those engineers.

Of course AMD won't e-mail me back and that is ok, but I see them allowing Intel to take 64bit and make it their own. AMD has plenty to be proud of and they should be letting everyone know. But as always I hear the ‘hammer' being pulled back, and another shot to take place, but this time it may be a shot to the heat.

I just want to say thank you to the engineers at AMD for all their hard work, regardless of what management and marketing does to it afterwards.

Thanks for letting me rant Mike,

Richard Getz

Overclocking the Athlon 2500+ mobile

Dear Editors,

I hate to burst anyone's bubble about the Athlon 2500+ Mobile CPU overclocking to 4000+, but that is not that incredible. Now, if you were talking about overclocking a standard desktop XP 2500+ on RA, that would be something.

I personally have recently overclocked a mobile XP 2500+ to 2.6GHz not only on RA, but at stock desktop voltage (1.7v). Check out the reviews for the Athlon XP 2500+ mobiles on Newegg.com, some have gone to 2.7GHz on RA with only a simple V-core voltage boost to 1.9v.

Those XP 2500+ mobiles are a fantastic CPU for overclocking, and you can buy one for about $100 USD.

Charles Truitt